The IBM data warehouse architecture

Reviewer: Kalman Balogh

This issue of the Communications of the ACM is dedicated to the rich and complicated theme of Data Warehousing. It is not an introduction to the theme, rather a state-of-art from practical point of view. It droves our attention to the process of Data Mart and Data Warehouse creation and usage before the big boom of its application. The series of articles emphasises and illustrates in more and more detail, why careful selection of the approach and the tools, building the projects and planning and design the implementation processes is inevitable. After the overview it starts with managerial considerations. It shows the possible ways, the costs, the intangible and measurable benefits. To clarify notions and purposes, it contrasts the transaction-oriented operational databases and the functional division or enterprise-wide information stores. It explains role of the information catalog or repository and metadata management. Some articles describe concrete developments at different large firms and organisations (logistics, investment, communications firm, health care). Successful architectures and some vendors (IBM, NCR) and development tools are introduced. It was not the aim to enumerate and compare the main vendors of the technology, but rich and interesting sites on the Web are enumerated.

Reviewer: William T. O'Connell

The authors combine an overview of a data warehouse architecture with a discussion of packaged solutions offered by IBM. Data warehouses are a complex integration of a variety of products. As decision support and analytical processing become more pervasive, the ability to transparently integrate both new and legacy data becomes imperative. Moreover, to complicate this integration, existing installations typically own data managed by several vendors (including nonrelational data). To ease the integration burden, the industry trend is to offer solutions and service support for enterprise-wide decision support systems. The paper first provides a short synopsis of data warehouse problems, topologies, and objectives. It then discusses relational database management system (RDBMS) warehouse support, middleware, metadata management, and packaged solutions. Finally, the authors address object/relational support and consider where IBM is going next. The authors' intention is to give an overview of the architectural infrastructure required to build and manage complex warehouses. This may include centralized or distributed warehouses (with or without data marts) and installations requiring the integration of multiple data stores. The authors focus on IBM solutions and products. They discuss a variety of IBM products, including the relational database family, middleware, replication tools, object support, metadata management, and nonrelational data support. Overall, the paper's ideas are well organized and well presented. Its primary audience is business analysts and users of data warehouses. I would recommend it to anyone who wants a high-level overview of data warehouse architectures focusing on IBM solutions and offerings. Those who want an insightful and comprehensive overview of the technical aspects of building a data warehouse should see Inmon [1].